Time-Frequency Analysis of a Moving Human Doppler Signature

摘要

In this report we analyze the Doppler signature of a moving human with the purpose of finding temporal change patterns that would allow detection and classification of movers. The analysis is based on computer models that simulate the operation of a pulse-Doppler radar. We start with an animated human mesh in walking motion and compute the frame-by-frame radar return using electromagnetic solvers such as the Finite Difference Time Domain (FDTD). The results are processed in the time-frequency domain via Short-Time Fourier Transforms (STFT) and displayed as spectrograms. In the end, we try to distinguish patterns in these spectrograms that are characteristic to a certain human motion type. The analysis emphasizes the phenomenological aspects of the Doppler signature, by looking at the relative contribution of various body parts and the effect of frequency, aspect angle, polarization and other radar parameters.